National
_....... I nstitute for
Commodity
Promotion
Research &
_~_ _~"""';"==.I Evaluation
PR~
Impact of Generic
Fluid Milk Advertising
on
Whole, Lowfat, and Skim
Milk Demand
by
Harry M. Kaiser
and
J. Carlos Roberte
Department of Agricultural, Resource, and Managerial Economics
College of Agriculture and Life Sciences
Cornell University, Ithaca, New York 14853
-
The National Institute For
Commodity Promotion Research And Evaluation
The National Institute for Commodity Promotion Research and
Evaluation was initially funded by a CSRS Special Grant in April 1994.
The Institute is an offshoot of The Committee on Commodity Promotion
Research (NEC63). A component of the Land Grant committee structure
to coordinate research in agriculture and related fields, NEC63 was
established in 1985 to foster quality research and dialogue on the
economics of commodity promotion.
The Institute's mission is to enhance the overall understanding of
economic and policy issues associated with commodity promotion
programs. An understanding of these issues is crucial to ensuring
continued authorization for domestic checkoff programs and to fund
export promotion programs. The Institute supports specific research
projects and facilitates collaboration among administrators and
researchers in government, universities, and commodity promotion
organizations. Through its sponsored research and compilations of
related research reports, the Institute serves as a centralized source of
knowledge and information about commodity promotion economics.
The Institute is housed in the Department of Agricultural, Resource, and
Managerial Economics at Cornell University in Ithaca, New York as a
component of the Cornell Program on Commodity Promotion Research.
Institute Objectives
• Support, coordinate, and conduct studies to identify key economic
relationships and assess the impact of domestic and export
commodity promotion programs on farmers, consumers, and the food
industry.
• Develop and maintain comprehensive data bases relating to
commodity promotion research and evaluation.
• Facilitate the coordination of multi-commodity and mUlti-country
research and evaluation efforts.
• Enhance both public and private policy maker's understanding of the
economics of commodity promotion programs.
• Facilitate the development of new theory and research methodology.
...
Table of Contents
Abstract
1
Preface
1
Introduction
1
The Model and Data
2
Estimation Procedures
3
The Regression Results
6
Policy Implications
9
Summary
9
References
10
-
Abstract
Preface
The purpose of this study was to determine whether
there is a statistical difference in sales responsiveness
to advertising among whole, lowfat, and skim milk
consumers. A case study for New York City which
used monthly time series demand data from 1986
through 1992 is presented. Separate per capita de­
mand functions were estimated for whole, lowfat, and
skim milk when per capita generic fluid milk adver­
tising expenditure was used as one of the explana­
tory variables. Other explanatory variables of per
capita sales included retail prices of whole, lowfat,
and skim milk, retail price of orange juice, per capita
income, and a health index representing consumer
concerns about fat in one's diet.
Harry M. Kaiser is an Associate Professor in the De­
partment of Agricultural, Resource, and Managerial
Economics at Cornell University, and Co-Director of
the National Institute for Commodity Promotion Re­
search and Evaluation (NICPRE). J. Carlos Roberte
is a Research Associate in the Department of Agricul­
tural, Resource, and Managerial Economics at Cornell
Unviersity. The authors thank Valerie Johnson for
her thorough editing and layout of this bulletin. The
authors would also like to thank Olan Forker and
Donald Liu for their helpful comments on an earlier
draft of this paper.
Long run sales responsiveness to generic milk adver­
tising were found to beO.14, 0.17, and 0.08 for whole,
lowfat, ad skim milk, respectively. These are higher
than previous estimates for generic fluid milk adver­
tising in New York City. The sales responsiveness
was found to be statistically significant at the 10%
level for whole and lowfat milk, but not significant for
skim milk. It was concluded that generic fluid milk
advertising, as currently structured, has had a posi­
tive and significant impact on whole and lowfat milk
demand, but little or no impact on skim milk de­
mand in the New York City market. It should be
noted that current generic milk advertising practice
does not distinguish among the three products.
The results suggest that the current message of the
fluid milk advertising campaign in New York City is
explicitly influencing actual and potential whole and
lowfat milk drinkers rather than skim milk consum­
ers. Therefore, it can be concluded that under cam­
paigns that do not differentiate among the three main
fluid milk products, it would be better to target actual
or potential consumers of whole and lowfat milk rather
than skim milk drinkers in New York City. It is clear
that any attempt to influence skim milk demand would
require a change in the current message. In addidtion,
since the sales responsiveness to advertising among
the three products are found to be different, future
research should study the separate impact of generic
fluid milk advertising on each fluid milk product. It
would be useful to apply this analytical approach to
other markets to determine whether similar conclu­
sions might hold, or whether the New York City mar­
ket is unique in its response to generic fluid milk ad­
vertising.
This is the second research bulletin published
by NICPRE. The mission of NICPRE is to enhance
the overall understanding of economic and policy is­
sues associated with commodity promotion programs.
An understanding of these issues is crucial to ensur­
ing continued authorization for domestic checkoff pro­
grams and to fund export promotion programs. The
bulletin will help program managers consider the im­
pacts of various allocation strategies used for promot­
ing different milk and dairy products. Future NICPRE
research bulletins will look at similar topics regarding
other agricultural commodities.
Introduction
Dairy farmers invest over $200 million annually in
programs designed to increase demand for fluid milk
and dairy products. These funds are collected by
assessing all farmers 15 cents per hundred pounds
on milk marketed in the continental United States.
Most of this investment is devoted to generic adver­
tising of milk and dairy products. Because of the sig­
nificant investment of money by producers, there have
been numerous studies on the effectiveness of ge­
neric milk advertising over the last 20 years. For in­
stance, Forker and Kinnucan summarized the results
of 47 studies of generic dairy advertising programs
throughout the world. Of these studies, the majority
(27 studies) have focused on generic fluid advertising
since fluid advertising represents the largest share of
advertising expenditures.
The impacts of generic fluid milk advertising
have been studied in a variety of geographic loca­
tions in the United States. Several studies have ex­
1
,
..
amined the impact of generic advertising on milk con­
sumption in New York City (Thompson and Eiler;
Kinnucan, 1986; Uu and Forker, 1988, 1990), in
Buffalo, New York (Kinnucan, 1987), and Roches­
ter, New York (Thompson, 1979). There have also
been investigations of advertising impacts of Califor­
nia (Thompson, 1974), of 12 federal milk marketing
orders in the United States (U.S. Department of Ag­
riculture), and at the national level (Ward and Dixon;
Uu et al.). While the magnitude of impacts have dif­
fered, all of these studies have found that generic ad­
vertising has increased consumption of fluid milk
above levels it would have been without advertising.
A common characteristic of all previous stud­
ies of generic fluid milk advertising has been the ag­
gregation of fluid milk products into a single product.
Since consumption patterns for some fluid products
have been quite different, aggregation of fluid milk
into one product results in a loss of useful informa­
tion on specific fluid milk product demand character­
istics. For example, per capita whole milk (3.5% fat
content) has steadily declined for decades, while per
capita consumption of lowfat (1% and 2% fat con­
tent) and skim milk has consistently increased over
time (see Figures 1, 2, and 3). Because of these dif­
fering trends, it would be useful to determine whether
whole, lowfat, and skim milk drinkers respond differ­
ently to the specific existing fluid milk advertising strat­
egy.
Accordingly, the purpose of this study was to
determine whether there is a statistical difference in
per capita sales response to generic fluid milk adver­
tising among whole, lowfat, and skim milk consum­
ers. A case study is presented for New York City us­
ing monthly time series demand data from 1986
through 1992. Separate demand functions are esti­
mated for whole, lowfat, and skim milk that include
generic fluid milk advertising expenditures as an in­
dependent variable in each equation. The policy
implications of the statistical results are then discussed.
This analysis provides information to help in fluid milk
targeting and allocation decisions. Generally, target­
ing the group with the largest sales responsiveness to
advertising will increase combined sales of fluid milk
products.
The Model and Data
FollOWing Uu and Forker (1988), generic advertising
was modeled as an extension of the consumer good
characteristics model by Lancaster.] In this model,
advertising may increase demand because it provides
consumers information on important attributes em­
bodied in the product. If advertising reduces infor­
mational search costs, the net result is a decrease in
the implicit price of the product characteristics, which
should increase consumption. Consequently, generic
advertising variables were included in the demand
functions for milk in addition to traditional demand
variables, e.g., own price, substitute prices, income,
etc.
The demand for each fluid milk product was
modeled in double-logarithmic form as follows:
3
(1)
In Q.
= U
+ L /3 ki In P kl + /3 41 In POJ, + /351 In INC,
k·]
J
3
j.Q
5']
l
+ L !PI.;' In H" + L 1t~ QD s + L YI.J1 In
AI;
+ u.'
j.Q
where subscript i = 1, 2, and 3 is for whole, lowfat,
and skim milk, respectively, and t = 1'00.,84 (monthly
data from January 1986 through December 1992).
The dependent variable in equation (1) was per capita
sales of the ith milk product in period t for the New
York City Metropolitan area. The independent vari­
ables included the retail price of each of the three
fluid products,2 retail price of orange juice (POJ,),
disposable per capita income (lNC I ), a health index
(HI) measuring the level of public concern about di­
etary fat in period t-j, seasonal intercept dummy vari­
ables (QD s ' where s = 1,2, and 3 for the first, second,
and third quarters, respectively), per capita generic
advertising expenditures (A,), and uti' an error term
with zero mean and variance S2.
,
For each fluid product, the retail price of the
other two fluid products and the retail price of orange
juice were used to capture the impact of substitutes
on demand, while disposable per capita income was
included to measure the effect of income on demand.
All prices and income variables were deflated. 3 The
IS ee Uu and Forker (1988) for a more detailed discussion
of extending Lancaster's consumer good characteristics
model to advertising.
2Since 1% and 2% milk were aggregated to represent lowfat
products in the model, the retail prices for each product
were aggregated by taking a weighted average of the two
prices, where the weights are equal to the relative market
shares of each product.
3S pecifically, the whole, lowfat, and skim milk price in the
whole milk and skim milk demand functions were deflated
by the dairy product price index for New York City. For the
lowfat demand equation, these prices were deflated by the
food and beverage price index for New York City. The
retail price of orange juice was deflated by the food and
beverage price index for New York City. Finally, per capita
income was deflated by the consumer price index for all
items for New York City.
2
­
retail price of orange juice used was that of the North­
east region of the United States. The health concern
index, which was constructed by Ward based on a
sample of consumers, gave the percent of consumers
expressing strong or moderate concern about fat in
their diets. This variable was included because con­
sumer concern about fat is expected to be an impor­
tant factor causing the decline in whole milk consump­
tion, and the increase in lowfat and skim milk con­
sumption. Since consumers' past concerns about fat
affect current consumption patterns, the health con­
cern index included current as well as lagged values.
Quarterly intercept dummy variables were included
to capture the impact of seasonality on demand.
under consideration, say Yf De-trending y, will be
appropriate if the DGP for the series is given by:
(2)
Y, = a o + a l T,
+ E"
where T, is a time trend, Et, is white noise, and aoand
a l are parameters. But if the DGP for Yt is given by:
I
(3)
Y, = a o + L Y,; + E"
;= t
then Yt should be differenced.
As discussed extensively in Banerjee et al. and
Davidson and MacKinnon, a unit root test can be used
to choose between de-trending and differencing. One
of these tests is the "augmented Dickey-Fuller" test
(ADFT) for unit roots. The ADFT is based on the
following regression equation:
To measure the impact of generic advertising
on the demand, monthly per capita generic fluid ad­
vertising expenditures were included for each fluid
milk product. Expenditures were deflated by the
media price index to adjust for inflation. Since ge­
neric fluid milk advertising has not been found to be
product specific (e.g., whole or skim milk), the same
level of milk advertising expenditures in the whole,
lowfat, and skim milk demand equations were in­
cluded. Current and lagged values of this variable
were included because it is well known that there are
carryover effects of advertising on demand. Includ­
ing current and lagged values of advertising is consis­
tent with the majority of past studies (e.g., Ward and
Dixon, Liu and Forker (1988), Liu et a1., Kaiser et
al.).
I
(4)
,i
Y,
=
a o + a l Y'_I + a z T, + L 0,
,iy._;
+ E"
i" 1
where ~ is the first-difference operator, ao' at' a 2 ,
and 8; are parameters, and Et is white noise. The test
statistic for the ADFT, say., is the t-ratio for the null
hypothesis that at =0. However,. does not follow a
t-distribution. Critical values for t are provided in
Fuller. The null hypothesis that Yt has a unit root (Le.,
y, is nonstationary) can be rejected if the calculated
value of. is smaller than the critical value correspond­
ing to the selected significance level. In such a case,
Yt must be differenced in order to achieve stationarity.
The ADFT was applied to all continuous vari­
ables in (l) using the COINT command in Shazam
version 7.0. The lag length I in (4) was set automati­
cally by the Shazam routine. For all the series, the
null hypothesis of nonstationarity could not be re­
jected at the 5% significance level. Based on this re­
sult, all variables were differenced once to make them
stationary. Since the first difference of the log of the
retail whole milk and skim milk prices were also found
to be nonstationary, these variables were differenced
twice.
Estimation Procedures
Figures 1 to 4 show that sales of whole milk and ad­
vertising expenditures have trended downward over
the sample period, while sales of lowfat and skim milk
trended upward. This suggests that these variables
may be nonstationary. Inferences based on standard
asymptotic theory may be misleading when data ex­
hibit nonstationary behavior (Davidson and
MacKinnon). The relationship between two variables
that exhibit a trend may appear to be significant when
in fact the only thing they have in common is the
trend. This phenomenon is referred to as "spurious
regression" in the econometrics literature.
The lag weights <Pt.ji and Y,.ji in (1) were approxi­
mated using third and second degree polynomials,
respectively. In addition, end point restrictions were
imposed on the distribution of the Y, _ coefficients.
The lag length for both generic advertising expendi­
tures and the health index was originally set to 12
months. The lag length in the final model specifica­
tion for each variable was then determined sequen­
tially using the following procedures. First, the num­
ber of lags for advertising expenditures was reduced
until the t-test for the significance of the last lag coef­
ficient could reject the null hypothesis of being equal
')1
Two methods are commonly used to correct
for this problem (Banerjee et al.): (l) de-trending
and (2) differencing the nonstationary series as many
times as needed to make them stationary. The ap­
propriateness of each method depends on the form
of the data generating process (DGP) for the series
3
­
14.0
12.0
T
•••
• •
.... •
•
••
• ••
I:
a"
s:: 10.0
0
•
••••
•
•
.... • • ••••••
• •
rJ)
I-<
(l)
0..
I-<
• •••• • • ..
.....
8.0
(l)
0..
rJ)
]
6.0
;:l
0
p..
4.0
2.0
0.0
1
13
25
37
49
Time (1986.01 = 1)
61
73
Figure 1. Per capita whole milk demand in New York City, 1986.01 - 1992.12.
45
4.0
3.5
s::
0rJ)
I-<
I-<
(l)
0..
rJ)
§
0
p..
1
_ ~_.~--.:"'-::a'· ••• •
3.0 •. rr.·
(l)
0..
•
• ••
2.5
2.0
1.5
1.0
0.5
0.0
1
13
25
37
49
Time (1986.01 = 1)
61
73
Figure 2. Per capita lowfat milk demand in New York City, 1986.01 - 1992.12.
4
­
1.6
1.4
c:
0
<Jl
I-<
(jJ
1.2
1.0
0..
I-<
(jJ
0..
0.8
<Jl
"'0
c:;::3 0.6
0
~
0.4
0.2
0.0
13
1
25
37
49
Time (1986.01
61
73
= 1)
Figure 3. Per capita skim milk demand in New York City, 1986.01 - 1992.12.
0.040
•
0.035
0.030
c:0 0.025
<Jl
I-<
(jJ
0..
I-<
(jJ
0..
fA
•
•• •
0.020 •
0.015
0.010
•
• •
•
••
• •
•
• •••
• • ••
•
•
• • ••
• • •••
•
•
•• ••• • •
•
•
•
•
•
•
•
•
0.005
•
0.000
1
13
25
37
49
•• • •
• •
•
•
•
•
•
• • • ••
•
... • • • •• ••
•
61
73
Time (1986.01 = 1)
Figure 4. Per capita generic fluid milk advertising in New York City, 1986.01 ­
1992.12.
5
-
Table 1. Estimated Whole Milk Demand Equation.
Variable
Estimated Coefficient
Constant
In (Deflated Whole Milk Price)
In (Deflated Lowfat Milk Price)
In (Deflated Orange Juice Price)
In (Deflated Per Capita Income)
In (Health Index),
In (Health Index)'_l
In (Health Index)'_2
In (Health Index)'_3
In (Health Index)'_4
In (Health Index)'_5
In (Health Index)'6
In (Health Index)'_7
In (Health Index)'_8
In (Health Index), 9
Sum of Lagged Health Index Coefficients
In (Deflated Generic Milk Advertising Expenditures),
In (Deflated Generic Milk Advertising Expenditures),l
In (Deflated Generic Milk Advertising Expenditures)'2
In (Deflated Generic Milk Advertising Expenditures), 3
In (Deflated Generic Milk Advertising Expenditures)'4
In (Deflated Generic Milk Advertising Expenditures), 5
In (Deflated Generic Milk Advertising Expenditures)'6
In (Deflated Generic Milk Advertising Expenditures)'7
In (Deflated Generic Milk Advertising Expenditures),_8
In (Deflated Generic Milk Advertising Expenditures),9
In (Deflated Generic Milk Advertising Expenditures)'l0
In (Deflated Generic Milk Advertising Expenditures),_ll
Sum of Lagged Advertising Coefficients
Quarter 1 Dummy Variable
Quarter 2 Dummy Variable
0.008
-0.051
0.244
0.209
0.915
-0.036
-0.054
-0.073
-0.091
-0.110
-0.129
-0.148
-0.168
-0.188
-0.208
-1.204
0.002
0.004
0.008
0.010
0.013
0.D15
0.017
0.018
0.018
0.016
0.013
0.008
0.140
-0.020
-0.022
The Regression Results
The statistical results for the three demand equations
are presented in Tables 1, 2 and 3. The final model
specification was achieved after dropping from the
original model those variables that had t-values less
than one in absolute value and/or signs inconsistent
with prior expectations. For each fluid milk product,
the joint null hypothesis that the coefficients on the
deleted variables are equal to zero was tested using
1.57
-0.21
1.04
2.04
11.61
-0.32
-0.55
-0.75
-0.92
-1.09
-1.30
-1.57
-1.89
-2.08
-1.94
-1.50
0.32
0.48
0.66
0.86
1.05
1.23
1.36
1.45
1.49
1.50
1.48
1.46
1.30
-2.07
-2.24
0.691
2.425
Adjusted R2
Durbin-Watson
F-Statistic for Significance of Skim Milk Price and
Quarter 3 Dummy Variable
to zero at the 10% level. 4 Then, the lag length for
the health index was determined following the same
procedure.
t-ratio
1.299
an F-test. The value of the F-statistic for each prod­
uct is reported at the bottom of each table. The final
model for whole milk demand did not include the
retail skim milk price and the intercept dummy vari­
able for the third quarter. The final model for lowfat
milk demand also did not include the retail skim milk
price and none of the intercept quarterly dummy vari­
ables. The final model for skim milk did not include
the retail whole milk price. The adjusted coefficients
~or generic milk advertising expenditures, a one-sided t­
test was used, based on the prior expectation that the im­
pact of this variable on sales is nonnegative. Similarly, a
nonpositive effect of the health index on whole milk sales
was expected. and, therefore a one-sided t-test was used to
select the lag length for this variable.
6
-
Table 2. Estimated Lowfat Milk Demand Equation.
Variable
Estimated Coefficient
Constant
In (Deflated Lowfat Milk Price)
In (Deflated Whole Milk Price)
In (Deflated Orange Juice Price)
In (Deflated Per Capita Income)
In (Health Index),
In (Health Index)t.]
In (Health Index)t.2
In (Health Index)t3
In (Health Index). 4
In (Health Index)"5
In (Health Index)t6
In (Health Index)'7
In (Health Index)'.B
In (Health Index)"9
Sum of Lagged Health Index Coefficients
In (Deflated Generic Milk Advertising Expenditures),
In (Deflated Generic Milk Advertising Expenditures)t-1
In (Deflated Generic Milk Advertising Expenditures),.2
In (Deflated Generic Milk Advertising Expenditures)•.3
In (Deflated Generic Milk Advertising Expenditures)'4
In (Deflated Generic Milk Advertising Expenditures), 5
In (Deflated Generic Milk Advertising Expenditures),.6
In (Deflated Generic Milk Advertising Expenditures),.7
In (Deflated Generic Milk Advertising Expenditures).-8
In (Deflated Generic Milk Advertising Expenditures),.9
In (Deflated Generic Milk Advertising Expenditures), 10
In (Deflated Generic Milk Advertising Expenditures)•. l l
Sum of Lagged Advertising Coefficients
0.008
-0.110
0.386
0.204
0.993
-0.083
-0.101
-0.119
-0.136
-0.153
-0.168
-0.184
-0.198
-0.212
-0.225
-1.578
0.002
0.005
0.008
0.012
0.015
0.018
0.021
0.022
0.022
0.0120
0.0160
0.009
0.168
Adjusted R2
Durbin-Watson
F-Statistic for Significance of Skim Milk Price and
Quarterly Dummy Variables
t-ratio
1.82
-0.32
1.73
1.89
11.77
-0.65
-0.96
-1.22
-1.39
-1.54
-1.72
-1.96
-2.21
-2.28
-2.00
-1.95
0.36
0.53
0.74
0.95
1.17
1.36
1.50
1.58
1.62
1.63
1.61
1.56
1.43
0.670
2.382
0.735
on the demand for whole and lowfat milk than the
own price. For instance, both the retail lowfat milk
and orange juice price had larger elasticities than the
own price in the whole milk demand equation. A
similar result holds for the lowfat milk demand equa­
tion. However, the own price elasticity of demand
for skim milk was larger than the two substitute prices
in the skim milk demand equation. Deflated per capita
income was highly significant in all three equations,
and ranged in value from .86 to .99. The estimated
income elasticities were twice as high as previous es­
timates for fluid milk demand in New York City, e.g.,
0.42 (Kinnucan, 1986), 0.41 (Kinnucan, Chang, and
Venkateswaran), 0.48 (Liu and Forker, 1988), and
0.36 (Liu and Forker, 1990). The health concern
index was significant in the whole and lowfat demand
equations, but not significant in the skim milk demand
of determination for the estimated whole, lowfat, and
skim milk demand functions were .69, .67, and .60,
respectively.
Not surprisingly, all of the estimated own price
elasticities were inelastic, -0.05 (whole milk), -0.11
(lowfat milk), and -0.33 (skim milk). In fact, only the
skim milk own price elasticity was significant at the
10% level. The skim milk price elasticity was consis­
tent with previous studies of New York City own fluid
milk price elasticities, e.g., -0.32 (Kinnucan, Chang,
and Venkateswaran), -0.29 (Liu and Forker, 1988),
and -0.18 (Liu and Forker, 1990). Alternatively, the
whole and lowfat price elasticities were similar to a
previous study of New York City by Kinnucan (1986),
who estimated an own fluid price elasticity of -0.095.
The price of substitutes had a more significant impact
7
-
Table 3. Estimated Skim Milk Demand Equation.
Variable
Estimated Coefficient
Constant
In (Deflated Skim Milk Price)
In (Deflated Lowfat Milk Price)
In (Deflated Orange Juice Price)
In (Deflated Per Capita Income)
In (Health Index),
In (Health Index)tl
In (Health Index)'_2
In (Health Index)t_3
In (Health Index)t_4
In (Health Index)tS
In (Health Index)t_6
In (Health Index)t_7
In (Health Index)t_8
In (Health Index)t9
In (Health Index)'lo
In (Health Index)'_ll
In (Health Index)'l2
Sum of Lagged Health Index Coefficients
In (Deflated Generic Milk Advertising Expenditures),
In (Deflated Generic Milk Advertising Expenditures)tl
In (Deflated Generic Milk Advertising Expenditures)t_2
In (Deflated Generic Milk Advertising Expenditures)t3
In (Deflated Generic Milk Advertising Expenditures)t_4
In (Deflated Generic Milk Advertising Expenditures)tS
In (Deflated Generic Milk Advertising Expenditures)t_6
In (Deflated Generic Milk Advertising Expenditures)t7
In (Deflated Generic Milk Advertising Expenditures)t_8
In (Deflated Generic Milk Advertising Expenditures),_9
In (Deflated Generic Milk Advertising Expenditures)'lo
In (Deflated Generic Milk Advertising Expenditures)tll
Sum of Lagged Advertising Coefficients
Quarter 1 Dummy Variable
Quarter 2 Dummy Variable
Quarter 3 Dummy Variable
0.018
-0.330
0.102
0.215
0.856
0.097
0.066
0.043
0.028
0.020
0.021
0.030
0.046
0.071
0.103
0.143
0.192
0.248
1.106
-0.006
-0.009
-0.008
-0.004
0.001
0.007
0.013
0.018
0.021
0.021
0.019
0.012
0.083
0.056
0.017
0.023
In the final estimated equations, generic fluid
advertising was lagged eleven months, indicating an
advertising carry-over effect of almost one year. The
long run generic advertising elasticities (sum of all co­
efficients) were 0.14,0.17, and 0.08 for whole, lowfat,
and skim milk, respectively. The advertising elastici­
-1.92
-1.35
0.38
1.82
8.94
0.71
0.57
0.39
0.25
0_18
0.18
0.25
0.39
0.62
0.96
1.39
1.79
2.02
0.93
-1.07
-0.87
-0.61
-0.30
0.06
0.44
0.83
1.17
1.46
1.68
1.84
1.96
0.64
4.48
1.34
1.74
0.600
2.166
0.373
Adjusted R2
Durbin-Watson
F-Statistic for Significance of Whole Milk Price
equation. The long run elasticities for the health in­
dex were -1.2, -1.6, and 1.1 for whole, lowfat, and
skim milk, respectively, indicating that health concerns
about fat have a negative effect on whole and lowfat
milk demand and positive (but not statistically signifi­
cant) impact on skim milk demand.
t-ratio
ties were higher than previous estimates for sales re­
sponsiveness to generic fluid milk advertising in New
York City, e.g., 0.051 (Kinnucan, 1986), 0.016
(Kinnucan, Chang, and Venkateswaran), 0.002 (Uu
and Forker, 1988), and 0.03 (Uu and Forker, 1990).
Based on a t-test, the long run generic advertising
elasticity was significant at the 10% level for whole
and lowfat milk, but not significant for skim milk. Thus,
the results indicate that existing and previous generic
fluid milk advertising has had a positive impact on
whole and lowfat milk demand, but has had no sig­
nificant impact on skim milk demand in the New York
City Market.
8
-
Policy Implications
There are several policy implications of these results.
First, the results indicate that generic fluid advertising
in New York City has been more effective in increas­
ing whole and low fat milk demand compared with
skim milk demand. This suggests that the message of
current and past fluid milk advertising campaigns may
have been implicitly targeted to actual or potential
consumers of whole and lowfat milk rather than skim
milk. Recall that current and past generic fluid milk
advertising did not differentiated among the three fluid
milk products. An alternative to the current advertis­
ing strategy of no differentiation among the three fluid
milk products might be to target actual or potential
consumers of whole and lowfat milk since they are
more responsive to past and present generic adver­
tising.
Second, given changes in consumption patterns
towards lower fat products, there might be some ad­
vantage to focusing on promoting lowfat milk prod­
ucts. The results of this study indicate that lowfat milk
drinkers (in New York City) respond positively to milk
advertising. Hence, targeting this group of milk con­
sumers would appear to be a good way of increasing
overall milk demand. The results also indicate that
targeting skim milk consumers might be a gamble.
This group of milk drinkers have been the least re­
sponsive to previous and current advertising. As was
argued above, the advertising message would have
to be altered under a skim milk advertising campaign.
Third, since the sales response to generic fluid
advertising among the three products were found to
be different, future research on generic fluid milk ad­
vertising should disaggregate milk products accord­
ingly whenever possible. While generic fluid adver­
tising did not have a significantly positive impact on
skim milk demand in the New York City market, this
result may be different in other markets. Hence, it
would be useful to apply this empirical approach to
other markets to determine whether New York City is
unique or representative of other locations. Certainly
the results from other markets should be obtained
before making any generalizations to larger geographic
locations.
Summary
pacts of advertising on specific products, i.e., whole,
lowfat, and skim milk. However, since consumption
patterns for these products are different, useful infor­
mation is lost in advertising evaluation studies when
looking only at the aggregate impact on the fluid milk
category. Consequently, the purpose of thiS study
was to determine whether there is a statistical differ­
ence in sales responsiveness of advertising among
whole, lowfat, and skim milk consumers. A case study
for New York City was presented using monthly time
series demand data from 1986 through 1992. Sepa­
rate demand functions were estimated for whole
lowfat, and skim milk when generic fluid milk adver~
tising expenditures were used as one of the explana­
tory variables.
The results indicated that the long run sales re­
sponsiveness to generic milk advertising is 0.14,0.17,
and 0.08 for whole, lowfat, and skim milk, respec­
tively. These are higher than previous estimates for
generic fluid milk advertising elasticities in New York
City. Based on a t-test, the long run sales response
to generic advertising was found to be significant at
the 10% level for whole and lowfat milk, but not sig­
nificant for skim milk. Thus, it was concluded that
generic fluid milk advertising, as currently structured,
has had a positive impact on whole and lowfat milk
demand, but little or no impact on skim milk demand
in New York City.
There are several implications of these results.
First, current and past message of fluid milk advertis­
ing campaigns is explicitly influencing actual and po­
tential whole and lowfat milk drinkers rather than skim
milk consumers. Therefore, under campaigns that
do not differentiate among the three main fluid milk
products, it would be better to target actual or poten­
tial consumers of whole and lowfat milk rather than
skim milk drinkers. Second, changing advertising
strategies to promote lowfat milk may lead to an in­
crease in overall consumption, but the same might
not be true for focusing advertising on skim milk. Thus,
any attempt to influence skim milk demand would
require a change in the current message. Third, since
the sales response to milk advertising among the three
products have been found to be different, future re­
search on generic fluid milk advertising should study
the impact on each fluid milk product separately. Fi­
nally, it would be useful to apply this approach to
other U.S. markets to determine whether similar re­
sults would be found, or whether New York City is
unique in its response to generic fluid milk advertis­
ing.
Previous research on the effectiveness of generic fluid
milk advertising has not identified the differential im­
9
-
References
Banerjee, A J.J. Dolado, J.w. Galbraith, and D.F. Hendry.
Co-Integration Error-Correction and the Economet­
ric Analysis of Non-Stationary Data. New York:
Oxford University Press, 1993.
DaVidson, R. and J.G. Mackinnon. Estimation and Infer­
ences in Econometrics. New York: Oxford Univer­
sity Press, 1993.
Forker, O.D. and H W Kinnucan. "Econometric Measure­
ment of Generic Advertising." International Dairy
Federation Special Issue No 9202. Brussels, Bel­
gium, 1991.
Fuller, WA Introduction to Statistical Time Series. New
York: John Wiley and Sons, 1976.
Kaiser, H.M., O.D. Forker, J. Lenz, and C.H Sun. "Evalu­
ating Generic Dairy Advertising Impacts on Retail,
Wholesale, and Farm Milk Markets." Journal of
Agricultural Economics Research. 44(1994):3-17.
Kinnucan, H.W, HS. Chang, and M. Venkateswaran.
"Generic Advertising Wearout." Review of Market­
ing and Agricultural Economics. 61(1993):401-15.
Kinnucan, HW "Effect of Canadian Advertising on Milk
Demand: The Case of the Buffalo, New York Mar­
ket." Canadian Journal of Agricultural Economics.
35(1987):181-96.
Kinnucan, H.W "Demographic versus Media Advertising
Effects on Milk Demand: The Case of the New York
City Market." Northeastern Journal of Agricultural
Economics. 15(1986): 66-74.
Liu, D.J. and O.D. Forker. "Generic FlUid Milk Advertis­
ing, Demand Expansion, and Supply Response: The
Case of New York City." American Journal of Agri­
cultural Economics. 70( 1988):229-36.
Liu, D.J. and O.D. Forker. "Optimal Control of Generic
Fluid Milk Advertising Expenditures." American
Journal of Agricultural Economics. 72 (1990): 1047­
55.
Liu, D.J., H.M. Kaiser, O.D. Forker, and TD. Mount. "An
Economic Model of the U.S. Generic Dai~ Adver­
tising Program Using an Industry Model.' ~
eastern Journal of Agricultural Economics
19(1990):37-48.
Thompson, S.R. "The Response of Milk Sales to Generic
Advertising and Producer Returns in the Rochester,
New York Market." AE. Staff Paper 79-26, De­
partment of Agricultural Economics, Cornell Univer­
sity, Ithaca, New York, 1979.
Thompson, S.R. "Sales Response to Generic Promotion
Efforts and Some Implications of Milk Advertising
on Economic Surplus." Journal of the Northeast­
ern Agricultural Economics Council. 3(1974):78­
90.
Thompson, S.R. and D.A. Eiler. "Producer Returns from
Increased Milk Advertising." American Journal of
Agricultural Economics. 57(1975):505-08.
U.S. Department of Agriculture. USDA Report to Con­
gress on the Dairy Promotion Program. 1985-94.
Ward, R.W and B.L. Dixon. "Effectiveness of Fluid Milk
Advertising Since the Dairy Tobacco Adjustment Act
of 1983." American Journal of Agricultural Eco­
~. 71(1989):730-40.
Ward, R.W Health Index on Dietary Concerns over Fat.
Department of Agricultural Economics, University
of Florida, Unpublished, 1995.
..
-
OTHER A.R.M.E. RESEARCH BULLETINS
No. 94-03
The Geographic Structure of Milk
Hauling Cost and Efficiencies in
New York State
Eric Erba
James Pratt
No. 94-04
Price Transmission Theory and
Applications to Agroindustry:
Annotated Bibliography
Lisa A. Schwartz
Lois Schertz Willett
An
No. 94-05
Decision Making in Membership
Organizations: A Study of Fourteen
U.S. Cooperatives
Brian M. Henehan
Bruce L. Anderson
No. 94-06
Identifying a Reduced Set of
Salient Attributes that Influence
Consumers' Choice Among Whole, Low­
Fat, and Skim Milk for Beverage Use
Heiko Miles
Steven J. Schwager
John E. Lenz
No. 94-07
Dairy Farm Management Business
Summary New York State 1993
Stuart F. smith
Wayne A. Knoblauch
Linda D. Putnam
No. 94-08
The Effects of Time-of-Use
Electricity Rates on New York Dairy
Farms
Mark Schenkel
Richard N. Boisvert
No. 95-01
supermarket Bakery Consumers
Attitudes, Preferences, Behaviors
Edward W., McLaughlin
Gerard Hawkes
Kristen Park
Debra Perosio
No. 95-02
The Economics of Converting
Conventionally Managed Eastern
Vineyards to organic Management
Practices
Gerald B. White
No. 95-03
Dairy Farm Management Business
Summary New York State 1994
Stuart F. smith
Wayne A. Knoblauch
Linda D. Putnam
-